Home > BrainVoyager > Functional Analysis > GLM: Single Study > How to use an empirical hemodynamic response within the GLM modeling in BrainVoyager

How to use an empirical hemodynamic response within the GLM modeling in BrainVoyager

BrainVoyager version: 22.4.4
Dataset used: sub-02_ses-01_task-Localizer_run-01_bold and sub-02_ses-01_task-Localizer_run-02_bold of the Localizer data of newbi4fmri (Jody Culham and Kevin Stubbs and Ethan Jackson and Rebekka Lagace Cusiac (2020). newbi4fmri2020 Localizer. OpenNeuro. [Dataset] doi: 10.18112/openneuro.ds003433.v1.0.1, accessed 31 January 2023)
Download: EHR default file

 

Within BrainVoyager, the user has several options to influence the HRF modeling of the GLM analysis. These options can be found on the “HRF” tab of the Single Study GLM options. One can either:

  • choose some of the default options including the “Boynton” and the “Two Gamma” HRF function,
  • influence the parameters of these functions,
  • shift the HRF functions or
  • use a so called “measured mean response” or “empirical hemodynamic response”

This document will explain the use of the option mentioned last. It allows to use predefined files (so called estimated hemodynamic response or *.ehr files) which are based on e.g. data extracted from a region of interest of the same participant in another functional run.

Here we will use a face/body/hand localizer dataset including with two functional runs per participants and session. The design contains four predictors called "Body", "Faces", "Hand" and "Scrambled". We can use a region defined on the basis of the analysis of the first functional run to extract a measured mean response (which can then be used to define the model for the analysis of the second run). First, we run a GLM analysis to detect activations specific to the “Body” condition in the first localizer run of participant 02. We define the GLM model (using the default HRF convolution).

We define three different contrasts and combine these in a conjunction analysis:

  1. [Body +1] ≠ [Face -1]
  2. [Body +1] ≠ [Hand -1]
  3. [Body +1] ≠ [Scrambled -1]

The resulting FDR-corrected statistical map will show only voxels that respond significantly stronger to body images as compared to face, hand and scrambled images:

To define the EHR format, we need:

  1. an event-related averaging (AVG) file
  2. a region of interest (VOI)

We open the event-related averaging dialog (Analysis -> Specify Event-Related Averaging...) and define and save an AVG file (of course this step can also be performed before the GLM analysis):

The AVG file covers the whole expected response curve of the HRF (31 volumes in total in this case) and additionally defines a time window for the baseline estimation (-2 to 0 volumes in the default case). The “expected response plot” shows a a preview of the plot created when using the AVG on the functional data. Now we define a region of interest to extract the functional data. In this case, we select the extrastriate area (EBA) in the right hemisphere of the participant. We display the timecourse of the region.

We open the options of the ROI signal time course by clicking into the time course:

We load the event-related average (AVG file) created before in the “Event-related averaging” field. The data of the to-be-created event-related averaging plot will serve as the basis for our new EHR file.

We click into the event-related averaging plot to display the plot options.

To extract the data of the event-related averaging plot, we can use the “Data Table…” button. The table can be saved as a simple text file that be used with different external tools.

From now on, we have to use a BV-external approach to create the actual EHR file / format. We open the table data created on the basis of the event-related average in any text editor. We only need the first column displaying the percent signal change values (the second
column, representing the standard error, can be neglected). We select and copy the first column of the file.

We open the EHR default file (“empirical_bold_response.ehr”) – which can be downloaded using this link or can be requested from the Brain Innovation support team via support@brainvoyager.com.

We paste the data of the AVG output table into the EHR file.

To finish this step, we have to adapt the header of the EHR format and save it with a new name.

old:

     

new:

We have 31 instead of 29 time points and the duration of events based on the protocol changes to 16 (the  body condition includes blocks of 16 and 17 TRs duration). The rest of the header stays the same in this case, but might differ in another example.

 

We switch back to BrainVoyager, open a MNI VMR and link the second run of the same participant and paradigm as used before. We open the "Single Study GLM Options" of the “Single Study GLM” dialog and switch to the "HRF" tab.

We load the previously created EHR file in the corresponding “Measured mean responses” field. We open the Single Study GLM main dialog and click the “Define Preds” button. Please note that in this case, the same EHR values will be automatically used for all four predictors. In a specific case, you might want to adapt this. One can clearly see the
difference to the standard HRF model used before.

The adapted model can be saved in a new SDM file.